silero-vad/silero-vad.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Jit example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:54:25.940761Z",
     "start_time": "2020-12-15T11:54:25.933842Z"
    }
   },
   "outputs": [],
   "source": [
    "# imports\n",
    "import glob\n",
    "import torch\n",
    "from IPython.display import Audio\n",
    "torch.set_num_threads(1)\n",
    "\n",
    "from utils import (init_jit_model, get_speech_ts,\n",
    "                  save_audio, read_audio, \n",
    "                  state_generator, single_audio_stream)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Full audio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:54:27.939388Z",
     "start_time": "2020-12-15T11:54:27.936636Z"
    }
   },
   "outputs": [],
   "source": [
    "def collect_speeches(tss, wav):\n",
    "    speech_chunks = []\n",
    "    for i in tss:\n",
    "        speech_chunks.append(wav[i['start']: i['end']])\n",
    "    return torch.cat(speech_chunks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:54:28.415177Z",
     "start_time": "2020-12-15T11:54:28.231677Z"
    }
   },
   "outputs": [],
   "source": [
    "model = init_jit_model('files/model.jit', 'cpu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:54:28.560822Z",
     "start_time": "2020-12-15T11:54:28.549811Z"
    }
   },
   "outputs": [],
   "source": [
    "wav = read_audio('files/en.wav')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:54:30.088721Z",
     "start_time": "2020-12-15T11:54:29.019358Z"
    }
   },
   "outputs": [],
   "source": [
    "speech_timestamps = get_speech_ts(wav, model, num_steps=4) # get speech timestamps from full audio file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:54:30.198484Z",
     "start_time": "2020-12-15T11:54:30.188311Z"
    }
   },
   "outputs": [],
   "source": [
    "speech_timestamps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:54:30.816893Z",
     "start_time": "2020-12-15T11:54:30.782667Z"
    }
   },
   "outputs": [],
   "source": [
    "save_audio('only_speech.wav', collect_speeches(speech_timestamps, wav), 16000) # merge all speech chunks to one audio\n",
    "Audio('only_speech.wav')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Single audio stream"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:54:31.886189Z",
     "start_time": "2020-12-15T11:54:31.572194Z"
    }
   },
   "outputs": [],
   "source": [
    "model = init_jit_model('files/model.jit', 'cpu')\n",
    "wav = 'files/en.wav'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:54:35.624279Z",
     "start_time": "2020-12-15T11:54:32.049532Z"
    }
   },
   "outputs": [],
   "source": [
    "for i in single_audio_stream(model, wav):\n",
    "    if i:\n",
    "        print(i)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Multiple audio stream"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:40:13.406225Z",
     "start_time": "2020-12-15T11:40:13.206354Z"
    }
   },
   "outputs": [],
   "source": [
    "model = init_jit_model('files/model.jit', 'cpu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:41:08.470917Z",
     "start_time": "2020-12-15T11:41:08.467369Z"
    }
   },
   "outputs": [],
   "source": [
    "audios_for_stream = glob.glob('files/*.wav')\n",
    "len(audios_for_stream) # total 4 audios"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:41:25.685356Z",
     "start_time": "2020-12-15T11:41:16.222672Z"
    }
   },
   "outputs": [],
   "source": [
    "for i in state_generator(model, audios_for_stream, audios_in_stream=2): # 2 audio stream\n",
    "    if i:\n",
    "        print(i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Onnx example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:55:45.597504Z",
     "start_time": "2020-12-15T11:55:45.582356Z"
    }
   },
   "outputs": [],
   "source": [
    "# imports\n",
    "import glob\n",
    "import torch\n",
    "from IPython.display import Audio\n",
    "torch.set_num_threads(1)\n",
    "import onnxruntime\n",
    "\n",
    "from utils import (get_speech_ts, save_audio, read_audio, \n",
    "                   state_generator, single_audio_stream)\n",
    "\n",
    "def init_onnx_model(model_path: str):\n",
    "    return onnxruntime.InferenceSession(model_path)\n",
    "\n",
    "def validate_onnx(model, inputs):\n",
    "    with torch.no_grad():\n",
    "        ort_inputs = {'input': inputs.cpu().numpy()}\n",
    "        outs = model.run(None, ort_inputs)\n",
    "        outs = [torch.Tensor(x) for x in outs]\n",
    "    return outs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Full audio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:55:56.874376Z",
     "start_time": "2020-12-15T11:55:56.782230Z"
    }
   },
   "outputs": [],
   "source": [
    "model = init_onnx_model('files/model.onnx')\n",
    "wav = read_audio('files/en.wav')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:56:12.159463Z",
     "start_time": "2020-12-15T11:56:11.446991Z"
    }
   },
   "outputs": [],
   "source": [
    "speech_timestamps = get_speech_ts(wav, model, num_steps=4, run_function=validate_onnx) # get speech timestamps from full audio file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:56:20.488863Z",
     "start_time": "2020-12-15T11:56:20.485485Z"
    }
   },
   "outputs": [],
   "source": [
    "speech_timestamps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:56:27.908128Z",
     "start_time": "2020-12-15T11:56:27.870978Z"
    }
   },
   "outputs": [],
   "source": [
    "save_audio('only_speech.wav', collect_speeches(speech_timestamps, wav), 16000) # merge all speech chunks to one audio\n",
    "Audio('only_speech.wav')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Single audio stream"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:58:09.012892Z",
     "start_time": "2020-12-15T11:58:08.940907Z"
    }
   },
   "outputs": [],
   "source": [
    "model = init_onnx_model('files/model.onnx')\n",
    "wav = 'files/en.wav'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:58:11.562186Z",
     "start_time": "2020-12-15T11:58:09.949825Z"
    }
   },
   "outputs": [],
   "source": [
    "for i in single_audio_stream(model, wav, run_function=validate_onnx):\n",
    "    if i:\n",
    "        print(i)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Multiple audio stream"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = init_onnx_model('files/model.onnx')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:59:09.381687Z",
     "start_time": "2020-12-15T11:59:09.378552Z"
    }
   },
   "outputs": [],
   "source": [
    "audios_for_stream = glob.glob('files/*.wav')\n",
    "len(audios_for_stream) # total 4 audios"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-12-15T11:59:27.712905Z",
     "start_time": "2020-12-15T11:59:21.608435Z"
    }
   },
   "outputs": [],
   "source": [
    "for i in state_generator(model, audios_for_stream, audios_in_stream=2, run_function=validate_onnx): # 2 audio stream\n",
    "    if i:\n",
    "        print(i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.3"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}